优化经验贝叶斯事故黑点识别与排序方法

Identification and ranking of accident black spots using advanced empirical Bayes method

为了对道路交通事故黑点进行准确有效识别和排序,提出基于事故统计数据及事故预测模型的优化经验贝叶斯黑点识别方法,并对该方法的工程实用性进行了优化。同时,从事故发生危险程度、安全治理可提升空间2个方面提出了优化经验贝叶斯黑点排序方法。选取2个城市的信号交叉口事故数据,利用基于事故统计数据的黑点识别方法对城市A的总事故黑点、伤亡事故黑点分别进行了识别,并按照事故发生危险程度进行了总事故黑点排序;利用基于事故预测模型的黑点识别方法对城市B的事故黑点进行识别,并分别按照事故发生危险程度、安全治理可提升空间2种排序规则进行了黑点排序。结果表明:在95%的显著性水平下,城市A共识别出24个总事故黑点及18个伤亡事故黑点;在99%的显著性水平下,城市B共识别出15个事故黑点;优化经验贝叶斯法在识别准确性、消除事故数随机波动及趋中心回归现象影响方面优于事故率法和质量控制法;基于安全治理可提升空间的排序指标倾向于筛选出安全治理投资收益比高的地点,基于事故发生危险程度的排序指标则倾向于筛选出与事故期望值偏离较大的地点。

To accurately and effectively identify and rank traffic accident black spots, an accident-data-based and an accident-prediction-model-based advanced empirical Bayes method was proposed. In addition, the practicability of the identification methods has been optimized. Furthermore, risk of accident involvement and safety benefits of treatment,two ranking criteria of black spots were suggested. Accident data for signalized intersections in two cities were collected. The accident-data-based advanced empirical Bayes method was used to identify total accident black spots and fatal and injury accident black spots in city A. The total number of accident black spots identified in city A were subsequently ranked according to the risk of accident involvement. Concurrently, black spots in city B were identified by the accident-prediction-model-based advanced empirical Bayes method, and were ranked according to risk of accident involvement and safety benefits of treatment two criteria. The results show that 24 total accident black spots and 18 fatal and injury accident black spots were identified in city A at 95% confidence level, while 15 black spots were identified in city B at 99% confidence level. The advanced empirical Bayes method is superior to the rate quality control and accident rate methods regarding accuracy, the ability to eliminate the impact of random fluctuation and the regression-to-the-mean effect of accident frequency. The ranking criterion of the safety benefits of treatment favors the locations that are more cost-effective to treat, while the ranking criterion of the risk of accident involvement favors the locations with higher deviation from the expected values. 7 tabs, 4 figs, 35 refs.